Design, synthesis, and computational affinity prediction of ester soft drugs as inhibitors of dihydrofolate reductase from Pneumocystis carinii

Article abstract of DOI:10.1016/j.ejps.2004.02.004

A series of dihydrofolate reductase (DHFR) inhibitors, where the methylenamino-bridge of non-classical inhibitors was replaced with an ester function, have been prepared as potential soft drugs intended for inhalation against Pneumocystis carinii pneumoni

Full text of DOI:10.1016/j.ejps.2004.02.004

European Journal of Pharmaceutical Sciences 22 (2004) 43–53
Design, synthesis, and computational affinity prediction of
ester soft drugs as inhibitors of dihydrofolate reductase
from Pneumocystis carinii
Malin Graffner-Nordberga,1, Karin Kolmodinb,2, Johan Åqvistb, Sherry F. Queenerc,
Anders Hallberga,∗
a
Department of Medicinal Chemistry, Uppsala Biomedical Center, Uppsala University, Box 574, SE-751 23 Uppsala, Sweden
b
Department of Cell and Molecular Biology, Uppsala Biomedical Center, Uppsala University, Box 596, SE-751 24 Uppsala, Sweden
c
Department of Pharmacology and Toxicology, School of Medicine, Indiana University, Indianapolis, IN 46202, USA
Received 13 October 2003; received in revised form 2 February 2004; accepted 17 February 2004
Abstract
A series of dihydrofolate reductase (DHFR) inhibitors, where the methylenamino-bridge of non-classical inhibitors was replaced with an
ester function, have been prepared as potential soft drugs intended for inhalation against Pneumocystis carinii pneumonia (PCP). Several of
the new ester-based inhibitors that should serve as good substrates for the ubiquitous esterases and possibly constitute safer alternatives to
metabolically stable DHFR inhibitors administered orally, were found to be potent inhibitors of P. carinii DHFR (pcDHFR). Although the
objectives of the present program is to achieve a favorable toxicity profile by applying the soft drug concept, a high preference for inhibition
of the fungal DHFR versus the mammalian DHFR is still desirable to suppress host toxicity at the site of administration. Compounds with a
slight preference for the fungal enzyme were identified. The selection of the target compounds for synthesis was partly guided by an automated
docking and scoring procedure as well as molecular dynamics simulations. The modest selectivity of the synthesized inhibitors was reasonably
well predicted, although a correct ranking of the relative affinities was not successful in all cases.
© 2004 Elsevier B.V. All rights reserved.
Keywords: Dihydrofolate reductase; Soft drug; Molecular dynamics; Pneumocystis carinii
1. Introduction
cause of mortality among patients suffering from AIDS
(Kovacs et al., 1984; Mills, 1986; Glatt and Chirgwin,
1990; McKenzie et al., 1991; Miller and Mitchell, 1995).
Current treatment of PCP with trimethoprim (TMP, Fig. 1),
a nonclassical inhibitor of dihydrofolate reductase (DHFR),
in combination with a sulfonamide (co-trimoxazole) is
still the first-line therapy in clinic (Fischl et al., 1988;
Walker and Masur, 1994; Miller and Mitchell, 1995;
Schliep and Yarrish, 1999). However, most often severe
side-effects associated with sulfa drugs lead to discontinu-
ation of therapy (Gordin et al., 1984; Glatt and Chirgwin,
1990; Masur, 1992; Roudier et al., 1994). Attempts have
been made to use drugs locally to treat PCP. Thus, inhaled
aerosolized pentamidine is used for prophylaxis (Golden
et al., 1989; Leoung et al., 1990; Monk and Benfield, 1990;
Hirschel et al., 1991), although this route of administration
is inadequate for treatment of the active infection. Systemic
The enhanced use of prophylaxis and highly active an-
tiretroviral therapy (HAART) in patients known to be at
risk for developing Pneumocystis carinii pneumonia (PCP),
has markedly decreased the incidence of PCP in patients
with AIDS (Schliep and Yarrish, 1999; Kaplan et al.,
2000; Arozullah et al., 2000). PCP is a serious pulmonary
disease caused by the fungus P. carinii and is the major
∗
Corresponding author. Tel.: +46-18-471-4284;
fax: +46-18-471-4976.
E-mail address: anders.hallberg@farmkemi.uu.se (A. Hallberg).
1
Present address: Department of Medicinal Chemistry, Biovitrum AB,
SE-751 82 Uppsala, Sweden.
2
Present address: EST Chemical Computing, AstraZeneca R&D
Mölndal, SE-431 83 Molndal, Sweden.
0928-0987/$ – see front matter © 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.ejps.2004.02.004

44
M. Graffner-Nordberg et al. / European Journal of Pharmaceutical Sciences 22 (2004) 43–53
OCH3
We have now prepared a series of lipophilic esters antic-
OCH3
ipated to serve as substrates for the ubiquitously distributed
esterases. We herein report the impact of various aromatic
scaffolds on the inhibition of DHFR from P. carinii DHFR
(pcDHFR) and the corresponding mammalian enzyme (rlD-
HFR). The selection of the target compounds for synthesis
was partly guided by molecular dynamics simulation and
the docking program AutoDock 3.0 (Morris et al., 1998).
NH2 CH3
NH2 CH3
OCH3
N
N
OCH3
N
N
H
H2N
N
H2N
N
N
TMQ
OCH3
O
PTX
NH2
N
NH2
N
OCH3
OCH3
N
O
H2N
H2N
OCH3
TMP
MGN103
2. Materials and methods
Fig. 1. Chemical structures of compounds TMQ, PTX, TMP and MGN103.
2.1. Chemistry: general procedures
administration of pentamidine exhibits a considerable tox-
icity (Pearson and Hewlett, 1985; Monk and Benfield,
1990; Walker and Masur, 1994; Schliep and Yarrish, 1999).
Trimetrexate (TMQ, Fig. 1) (Elslager and Davoll, 1974;
Elslager et al., 1983) and piritrexim (PTX, Fig. 1) (Grivsky
et al., 1980), are two new lipophilic agents originally devel-
oped as anticancer agents (Bertino et al., 1979; Duch et al.,
1982; Weir et al., 1982; Lin et al., 1987; Kovacs et al.,
1988). Recently, both drugs, TMQ in particular, have found
a place as second-line therapy in moderate to severe PCP
(Anonymous, 1994). Although TMQ and PTX are potent
inhibitors of DHFR from P. carinii, they are not selective
and inhibit the mammalian enzyme even more efficiently
(Allegra et al., 1987b; Broughton and Queener, 1991). The
clinical use of TMQ and PTX is therefore limited as a result
of their systemic host toxicity and an expensive co-therapy
with the rescue agent leucovorin (5-formyl-tetrahydrofolate)
is required (Allegra et al., 1987a,b; Falloon et al., 1990;
Sattler et al., 1990; Masur et al., 1993). Consequently, ef-
forts focus on finding more selective as well as more potent
antifolates.
The soft drug concept could serve as a new strategy
in the search for safer DHFR inhibitors with potential
use in the therapy against PCP. Soft drugs are active
isosteric–isoelectronic analogues of lead compounds that are
deactivated in a predictable and controllable way after ex-
erting their biological effects (Bodor and Buchwald, 2000).
These drugs are intended to undergo a fast metabolism to
inactive and non-toxic metabolites. Soft drugs are often
used for local treatment and are, in general, administered
near the site of action (Bodor and Buchwald, 2000). The
concept has successfully been utilized, for example, in the
design of the recently launched inhaled anti-asthma gluco-
corticosteroid budesonide (Clissold and Heel, 1984) that,
essentially, is devoid of systemic side effects in man due
to its rapid first pass metabolism in the liver (Johansson
et al., 1982). Furthermore, in an initial report, we recently
demonstrated with one example that a lipophilic ester acting
as a DHFR inhibitor was rapidly inactivated after oral or
intravenous administration in rat (Graffner-Nordberg et al.,
2000). We believe that the application of the soft drug con-
cept should allow for identification of new DHFR inhibitors
suitable for inhalation against PCP.
1H and 13C NMR spectra were recorded on a JEOL
JNM-EX270 spectrometer at 270 and 67.8 MHz, respec-
tively, and a JEOL JNM-EX400 spectrometer at 400 and
100 MHz, respectively. Thin-layer chromatography (TLC)
was performed by using aluminium sheets precoated with
silica gel 60 F254 (0.2 mm) type E; Merck. Chromatographic
spots were visualized by UV light. Column chromatography
was conducted on silica gel 60 (0.040–0.063 mm; Merck),
unless otherwise noted. Melting points (uncorrected) were
determined in open glass capillaries on an Electrothermal ap-
paratus. The Quest 210 organic synthesizer (Argonaut Tech-
nologies) was used in the syntheses of 3–4, 6–8, 12, 14–15.
In the syntheses of 10 and 11 microwave heating was per-
formed in a Smith SynthesizerTM single mode microwave
cavity producing continuous irradiation at 2450 MHz (Per-
sonal Chemistry AB, Uppsala, Sweden). The syntheses were
performed in heavy-walled glass Smith Process Vials sealed
with aluminum crimp caps fitted with a silicon septum. The
inner diameter of the vial filled to the height of 2 cm was
1.3 cm. Reaction mixtures were stirred with a magnetic stir
bar during the irradiation. The temperature, pressure and ir-
radiation power was monitored during the course of the reac-
tion. The average pressure during the reaction was 3–4 bar.
After completed irradiation, the reaction tube was cooled
with high-pressure air until the temperature had fallen be-
low 39 ◦C (ca. 2 min). The microwave irradiations were per-
formed under controlled conditions that make the procedure
highly safe, reliable and reproducible. Single mode irradi-
ation with monitoring of temperature, pressure and irradia-
tion power versus time was used throughout. The reaction
temperature was kept constant throughout the reaction in the
single mode cavity by an automatic power control. Caution!
When carrying out reactions with microwave irradiation in
closed vessels, extra caution is advisable. In addition to the
high pressure generated by the vapor pressure of volatile
components, the metal catalysts might precipitate and cause
a “thermal runaway”, increasing the pressure further. There-
fore, the use of special heavy-walled process vials is highly
recommended. SpeedVac® Plus SC250DDA (Savant) was
used in the evaporations of DMF. The elemental analyses
were performed by Mikro Kemi AB, Uppsala, Sweden and
were within 0.4% of the calculated values. The synthesis of

M. Graffner-Nordberg et al. / European Journal of Pharmaceutical Sciences 22 (2004) 43–53
45
2,4-diamino-6-bromomethylquinazoline has been described
elsewhere (Graffner-Nordberg et al., 2000). All commercial
chemicals were used without further purification.
the crude ester was purified as described for compound 1
yielding 34 mg (22% over two steps): 1H NMR (DMSO-d6)
δ 8.95–8.87 (m, 3H, ArH), 8.54 (s, 1H, ArH), 8.19–8.16 (m,
3H, ArH), 7.56–7.68 (m, 4H, H-7 + ArH), 7.35 (br s, 2H,
NH2), 7.25 (d, J = 8.58 Hz, 1H, H-8), 6.07 (br s, 2H, NH2),
5.46 (s, 2H, CH2); 13C NMR (DMSO-d6) δ 166.84, 162.50,
161.13, 152.67, 133.23, 131.72, 131.40, 131.21, 130.05,
129.54, 129.45, 128.18, 127.60, 127.51, 127.35, 127.08,
125.99, 124.60, 124.32, 123.47, 123.00, 109.98, 67.07 (one
carbon missing). Anal. (C24H18N4O2·1H2O) C, H, N.
2.1.1. (2,4-Diaminoquinazoline-6-yl)methyl 1-naphthoate
(1)
A solution of 2,4-diamino-6-bromomethylquinazoline
(0.53 mmol) in anhydrous DMF (2 ml) was added dropwise
to a mixture of 1-naphthoic acid (272 mg, 1.58 mmol), and
potassium carbonate (218 mg, 1.58 mmol) in DMF (total
amount: 5 ml). The reaction mixture was stirred under nitro-
gen at room temperature for 2 days before the crude product
was evaporated under reduced pressure on a small portion
of silica gel. The silica plug was loaded on the top of a silica
column, previously packed with dichloromethane, and the
crude ester was purified by repeated flash chromatography
[CH2Cl2:MeOH+NH3 (19:1 yielding 80 mg (44% over two
2.1.4. (2,4-Diaminoquinazoline-6-yl)methyl
9-anthracencarboxylate (4)
Compound 4 was prepared as described for 3 us-
ing the Quest 210 organic synthesizer starting with
9-anthracenecarboxylic acid (351 mg, 1.58 mmol). The re-
action was allowed to proceed for 2 days before it was
filtered and worked-up as described above for 3 yield-
ing 50 mg (24% over two steps): 1H NMR (DMSO-d6)
δ 8.95–8.87 (m, 3H, ArH), 8.54 (s, 1H, ArH), 8.19–8.16
(m, 3H, ArH), 7.56–7.68 (m, 4H, H-7 + ArH), 7.35 (br s,
2H, NH2), 7.25 (d, J = 8.58 Hz, 1H, H-8), 6.07 (br s, 2H,
NH2), 5.46 (s, 2H, CH2); 13C NMR (DMSO-d6) δ 168.62,
162.47, 161.12, 152.70, 133.56, 130.40, 129.21, 128.67
(2C), 127.42 (3C), 126.61, 125.78 (2C), 124.94, 124.54,
124.43 (2C), 109.90, 67.78 (three carbons missing). Anal.
(C24H18N4O2·0.25H2O) C, H, N.
1
steps): H NMR (DMF-d6) δ 8.88–8.84 (app d, 1H, ArH),
8.26 (d, J = 1.98 Hz, 1H, H-5), 8.22–8.18 (m, 2H, ArH),
8.05–8.02 (m, 1H, ArH), 7.71 (dd, J = 8.58, 1.98 Hz, 1H,
H-7), 7.68–7.55 (m, 3H, ArH), 7.46 (br s, 2H, NH2), 7.28
(d, J = 8.58 Hz, 1H, H-8), 6.12 (br s, 2H, NH2), 5.45 (s,
2H, CH2); 13C NMR (DMF-d6) δ 167.61, 163.80, 162.90,
153.92, 134.60, 134.19, 133.78, 131.71, 130.90, 129.48,
128.51, 128.44, 127.63, 127.07, 126.01, 125.76, 125.57,
124.81, 111.01, 67.64. Anal. (C20H16N4O2·0.4H2O) C,
H, N.
2.1.5. (2,4-Diaminoquinazoline-6-yl)methyl
2.1.2. (2,4-Diaminoquinazoline-6-yl)methyl
diphenylacetate (5)
2-naphthoate (2)
Compound 5 was prepared as described for 1 using
diphenylacetic acid (335 mg, 1.58 mmol). The reaction was
allowed to proceed for 2 days before it was filtered and
worked-up as described above for 1 yielding 72 mg (36%
over two steps): 1H NMR (DMF-d6) δ 8.08 (d, J = 1.65 Hz,
1H, H-5), 7.45 (dd, J = 8.58, 1.98 Hz, 1H, H-7), 7.41 (br s,
2H, NH2), 7.37–7.19 (m, 10H, ArH), 7.19 (d, J = 8.58 Hz,
1H, H-8), 6.16 (br s, 2H, NH2), 5.25 (s, 1H, CH), 5.17 (s,
2H, CH2); 13C NMR (DMF-d6) δ 172.79, 163.73, 153.51,
140.07 (2C), 133.71, 129.25 (2C), 129.18 (2C), 128.21,
127.78 (2C), 125.44 (2C), 124.87, 110.78, 67.55, 56.93
(2C) (three carbons missing). Anal. (C23H20N4O2) C, H, N.
Compound 2 was prepared as described for 1 using
2-naphthoic acid (272 mg, 1.58 mmol). The reaction was
allowed to proceed for 2 days before it was filtered and
worked-up as described above for 1 yielding 60 mg (33%
over two steps): 1H NMR (DMF-d6) δ 8.68 (app s, 1H,
ArH), 8.24 (d, J = 1.65 Hz, 1H, H-5), 8.14–8.11 (m,
1H, ArH), 8.00–7.98 (m, 2H, ArH), 7.70–7.57 (m, 4H,
H-7, ArH), 7.45 (br s, 2H, NH2), 7.27 (d, J = 8.58 Hz,
1H, H-8), 6.11 (br s, 2H, NH2), 5.41 (s, 2H, CH2); 13C
NMR (DMF-d6) δ 166.73, 163.80, 162.90, 153.88, 136.22,
133.63, 133.20, 131.46, 130.08, 129.27, 129.09, 128.46,
128.14, 127.67, 125.75, 125.66, 124.62, 111.00, 67.55 (one
carbon missing). Anal. (C20H16N4O2·0.25H2O) C, H, N.
2.1.6. (2,4-Diaminoquinazoline-6-yl)methyl
2ꢀ-benzylbenzoate (6)
2.1.3. (2,4-Diaminoquinazoline-6-yl)methyl
Compound 6 was prepared as described for 3 using the
Quest 210 organic synthesizer starting with 2-benzylbenzoic
acid (334 mg, 1.58 mmol). The reaction was allowed to
proceed for 2 days before it was filtered and worked-up
as described above for 3 yielding 53 mg (26% over two
9-phenanthrenecarboxylate (3)
Compound 3 was prepared using the Quest 210 organic
synthesizer starting with 2,4-diamino-6-bromomethylqui-
nazoline (0.53 mmol), 9ꢀ-phenanthrenecarboxylic acid
(351 mg, 1.58 mmol), and potassium carbonate (218 mg,
1.58 mmol) in DMF (total amount: 5 ml). The reaction was
allowed to proceed for 2 days before it was filtered and
rinsed with DMF. A small portion of silica gel was added to
the reaction mixture and the solvent was evaporated using
SpeedVac®. The silica plug was loaded on the top of a sil-
ica column, previously packed with dichloromethane, and
1
steps): H NMR (DMSO-d6) δ 8.06 (d, J = 1.65 Hz, 1H,
H-5), 7.83–7.80 (m, 1H, ArH), 7.54–7.04 (m, 12H, NH2
+ ArH), 6.04 (br s, 2H, NH2), 5.24 (s, 2H, CH2), 4.28 (s,
2H, CH2); 13C NMR (DMSO-d6) δ 166.81, 162.26, 160.83,
152.27, 141.39, 140.59, 132.96, 132.01, 131.32, 130.03,
129.51, 128.36 (2C), 128.05 (2C), 126.85, 126.31, 125.66,

46
M. Graffner-Nordberg et al. / European Journal of Pharmaceutical Sciences 22 (2004) 43–53
124.28, 124.06, 109.65, 64.48 (one carbon missing). Anal.
(C23H20N4O2) C, H, N.
charged with compound 9 (50 mg, 0.12 mmol) in DMF
(800 ␮l), phenylboronic acid (72.5 mg, 0.59 mmol) in
DMF (200 ␮l), PdCl2(PPh3)2 (0.8 mg, 1.2 mmol), and 2 M
Na2CO3 (150 ␮l), in 2 ml of DME:H2O:EtOH (7:3:2). The
reaction mixture was exposed to microwave irradiation for
120 s at 150 ◦C. The mixture was filtered through a sy-
ringe equipped with a Titan® filter (pore size 0.45 ␮m).
A small portion of silica gel was added to the solution.
The solvent was evaporated using SpeedVac®. Purifications
were conducted using flash chromatography as described
for compound 1 finally yielding 15 mg (34%) of the tar-
2.1.7. (2,4-Diaminoquinazoline-6-yl)methyl
2ꢀ-phenylbenzoate (7)
Compound 7 was prepared as described for 3 using the
Quest 210 organic synthesizer starting with 2-phenylbenzoic
acid (313 mg, 1.58 mmol). The reaction was allowed to
proceed for 2 days before it was worked-up as described
above for 3 yielding 47 mg (24% over two steps): 1H NMR
(DMSO-d6) δ 7.90 (app s, 1H, H-5), 7.78–7.09 (m, 13H,
ArH), 6.04 (br s, 2H, NH2), 5.05 (s, 2H, CH2); 13C NMR
(DMSO-d6) δ 168.00, 162.39, 162.11, 161.68, 161.05,
152.55, 141.23, 140.35, 132.87, 131.45, 130.93, 130.48,
129.28, 128.38, 128.13, 127.43, 127.20, 126.46, 124.32,
124.09, 109.70, 66.81. Anal. (C22H18N4O2·0.25H2O) C,
H, N.
1
get molecule: H NMR (DMSO-d6) δ 8.20–8.18 (m, 1H,
ArH), 8.10 (d, J = 1.65 Hz, 1H, H-5), 8.00–7.93 (m,
2H, ArH), 7.70–7.37 (m, 9H, ArH + NH2), 7.22 (d, J =
8.58 Hz, 1H, H-8), 6.07 (br s, 2H, NH2), 5.36 (s, 2H, CH2);
13C NMR (DMSO-d6) δ 165.98, 162.82, 161.22, 152.54,
141.08, 139.41, 133.46, 132.04, 132.73, 129.94, 129.49
(2C), 128.61, 128.38, 127.68, 127.55, 127.19 (2C), 124.71,
124.37, 110.21, 67.09. Anal. (C22H18N4O2·0.25H2O) C,
H, N.
2.1.8. (2,4-Diaminoquinazoline-6-yl)methyl
4ꢀ-phenylbenzoate (8)
Compound 8 was prepared as described for 3 using the
Quest 210 organic synthesizer starting with 4-phenylbenzoic
acid (313 mg, 1.58 mmol). The reaction was interrupted
after 2 days and further worked-up as described above
for 3 yielding 25 mg (13% over two steps): 1H NMR
(DMSO-d6) δ 8.11–7.21 (m, 14H, ArH), 6.12 (br s,
2H, NH2), 5.34 (s, 2H, CH2); 13C NMR (DMSO-d6) δ
165.48, 162.46, 160.81, 152.09, 144.76, 138.80, 133.03,
129.94, 129.11, 128.46, 128.21, 128.14, 127.30, 126.99,
124.28, 124.03, 109.89, 66.62 (four carbons missing). Anal.
(C22H18N4O2·0.75H2O) C, H, N.
2.1.11. (2,4-Diaminoquinazoline-6-yl)methyl
[2ꢀ,6ꢀ-dimethoxy]-3ꢀ-phenylbenzoate (11)
Compound 11 was performed as for compound 10 us-
ing 2,6-dimethoxyphenylboronic acid (108 mg, 0.59 mmol).
Work-up was conducted as described for the synthesis
of 10 providing 20 mg (39%) of the target molecule: 1H
NMR (DMSO-d6) δ 8.09 (app s, 1H, H-5), 7.92–7.89
(m, 1H, ArH), 7.78 (m, 1H, ArH), 7.60 (dd, J
8.58, 1.98 Hz, 1H, H-7), 7.55–7.46 (m, 2H, ArH), 7.35–7.30
(m, 3H, ArH + NH2), 7.22 (d, J 8.58 Hz, 1H,
H-8), 6.75 (d, J 8.58 Hz, 1H, H-8), 6.06 (br s,
=
=
=
2.1.9. (2,4-Diaminoquinazoline-6-yl)methyl
2H, NH2), 5.32 (s, 2H, CH2) 3.64 (s, 6H, CH3); 13C
NMR (DMSO-d6) δ 166.11, 162.80, 161.20, 157.32,
152.52, 136.21, 135.08, 133.50, 131.77, 129.89, 129.54,
128.54, 127.85, 127.71, 124.67, 124.46, 117.61, 110.19,
104.69, 66.93, 56.04 (2C) (two carbons missing). Anal.
(C24H22N4O4·2H2O) C, H; N: calcd, 12.0; found,
11.3.
3ꢀ-iodobenzoate (9)
Compound 9 was prepared as described for 1 starting
with 2,4-diamino-6-bromomethylquinazoline (2.52 mmol),
3-iodobenzoic acid (1.87 g, 7.57 mmol), and potassium car-
bonate (1.04 g, 7.57 mmol) in DMF (total amount: 20 ml).
The reaction was allowed to proceed for 2 days before it
was filtered and worked-up as described above for 1 yield-
ing 441 mg (42% over two steps). Recrystallization from
methanol afforded 360 mg (34%) as white needles: 1H NMR
(DMSO-d6) δ 8.25 (dd, J = 1.32, 1.32 Hz, 1H, ArH), 8.08
(d, J = 1.65 Hz, 1H, H-5), 8.04–7.98 (m, 2H, ArH), 7.59
(dd, J = 8.58, 1.65 Hz, 1H, H-7), 7.32 (br s, 2H, NH2), 7.34
(dd, J = 7.92, 7.92 Hz, 1H, ArH), 7.20 (d, J = 8.58 Hz,
1H, H-8), 6.05 (br s, 2H, NH2), 5.32 (s, 2H, CH2); 13C
NMR (DMSO-d6) δ 164.00, 162.12, 160.77, 152.34, 141.56,
137.03, 132.65, 131.34, 130.64, 128.28, 126.49, 124.22,
123.76, 109.58, 94.56, 66.69. Anal. (C16H13IN4O2) C,
H, N.
2.1.12. (2,4-Diaminoquinazoline-6-yl)methyl
3ꢀ,5ꢀ-dimethylbenzoate (12)
Compound 12 was prepared as described for 3 us-
ing the Quest 210 organic synthesizer starting with
3,5-dimethylbenzoic acid (237 mg, 1.58 mmol). The re-
action was allowed to proceed for 2 days before it was
filtered and worked-up as described above for 3 yielding
1
32 mg (19% over two steps): H NMR (DMSO-d6) δ 8.08
(d, J = 1.65 Hz, 1H, H-5), 7.32 (br s, 2H, NH2), 7.20 (d,
J = 8.58 Hz, 1H, H-8), 7.28 (s, 1H, ArH), 7.59–7.56 (m,
3H, NH2 + ArH), 6.06 (br s, 2H, NH2), 5.29 (s, 2H, CH2),
3.36 (s, 6H); 13C NMR (DMSO-d6) δ 165.87, 162.44,
161.09, 152.63, 138.02 (2C), 134.74, 133.08, 129.61,
127.10, 126.85 (2C), 124.57, 124.19, 109.90, 66.58, 20.68
(2C). Anal. (C18H18N4O2·1.75H2O) C, H; N: calcd, 16.12;
found, 15.50.
2.1.10. (2,4-Diaminoquinazoline-6-yl)methyl
3ꢀ-phenylbenzoate (10)
A heavy-walled glass Smith Process Vial sealed with
aluminum crimp caps fitted with a silicon septum was

M. Graffner-Nordberg et al. / European Journal of Pharmaceutical Sciences 22 (2004) 43–53
47
2H, NH2), 5.27 (s, 2H, CH2); 13C NMR (DMSO-d6) δ
162.34, 160.81, 157.74, 152.19, 147.61, 143.58, 133.11,
126.84, 124.27, 124.14, 118.61, 112.28, 109.71, 66.22. Anal.
(C14H12N4O3·0.75H2O) C, H, N.
2.1.13. (2,4-Diaminoquinazoline-6-yl)methyl
2-thiophenecarboxylate (13)
Compound 13 was prepared as described for 1 using
2-thiophenecarboxylic acid (202 mg, 1.58 mmol). The reac-
tion was allowed to proceed for 2 days before it was fil-
tered and worked-up as described above for 1 yielding 63 mg
2.2. Molecular dynamics
1
(40% over two steps): H NMR (DMF-d6) δ 8.28 (d, J =
1.65 Hz, 1H, H-5), 7.98 (br s, 2H, NH2), 7.96 (m, 1H, ArH),
7.85–7.83 (m, 1H, ArH), 7.72 (dd, J = 8.58, 1.98 Hz, 1H,
H-7), 7.36 (d, J = 8.58 Hz, 1H, H-8), 7.24–7.21 (m, 1H,
ArH), 6.81 (br s, 2H, NH2), 5.35 (s, 2H, CH2); 13C NMR
(DMF-d6) δ 163.99, 160.37, 149.38, 134.54 (2C), 129.85,
128.99, 125.03, 123.15, 110.71, 67.21 (three carbons miss-
ing). Anal. (C14H12N4O2S·0.5H2O) C, H, N.
The linear interaction energy (LIE) method (Åqvist et al.,
1994; Marelius et al., 1998c; Graffner-Nordberg et al., 2001)
was used to estimate the free energy of binding for a num-
ber of inhibitors, docked into the active sites of hDHFR
and pcDHFR, respectively, using AutoDock. The complexes
were further subjected to molecular dynamics simulations
from which the energy averages were calculated. All MD
simulations were performed using the software package Q
(Marelius et al., 1998b) and the GROMOS87 force field (van
Gunsteren and Berendsen, 1987). Non-standard charges of
NADPH and the inhibitors were derived as in earlier work
(Marelius et al., 1998a).
The starting coordinates for the simulations of DHFR
from P. carinii were taken from the crystal structure of the
ternary complex with the enzyme in complex with NADPH
and TMP (PDB entry 1DYR). The structure of human
DHFR was the L22Y mutant enzyme in complex with
NADPH and methotrexate (PDB entry 1DLS) from which
the wild type enzyme was modeled prior to simulation. The
protein-inhibitor complex was soaked in an 18 Å spherical
grid of SPC water molecules. Water molecules generated
closer than 2.4 Å to protein or inhibitor atoms were re-
moved. Ionic groups on the protein within about 10 Å from
the inhibitors were included, whereas those far from the
active site were modeled as neutral dipolar groups. The net
charge of the protein and cofactor was zero. Simulations of
the free inhibitors were carried out in water spheres of the
same size as the protein simulations. The water molecules
were subjected to radial and polarization surface restraints
and the local reaction field (LRF) (Lee and Warshel, 1992)
was used for handling long range electrostatics. The system
was heated to 300 K by increasing the temperature incre-
mentally during a 30 ps equilibration phase keeping the
protein atoms restrained to their crystallographic positions.
The system was then equilibrated without restraining the
protein in the simulation sphere for 50 ps. Data collection
was performed at 300 K using a time step of 1.5 fs.
2.1.14. (2,4-Diaminoquinazoline-6-yl)methyl
3-thiophenecarboxylate (14)
Compound 14 was prepared as described for 3 us-
ing the Quest 210 organic synthesizer starting with
3ꢀ-thiophenecarboxylic acid (202 mg, 1.58 mmol). The sil-
ica plug was loaded on the top of a silica column, previously
packed with dichloromethane, and the crude ester was pu-
rified as described for compound 1 yielding 34 mg (22%
over two steps): 1H NMR (DMSO-d6) δ 8.40–8.38 (m, 1H,
ArH), 8.06 (d, J = 1.65 Hz, 1H, H-5), 7.67–7.64 (m, 1H,
ArH), 7.56 (dd, J = 8.58, 1.98 Hz, 1H, H-7), 7.50–7.48 (m,
1H, ArH), 7.31 (br s, 2H, NH2), 7.20 (d, J = 8.58 Hz, 1H,
H-8), 6.05 (br s, 2H, NH2), 5.26 (s, 2H, CH2); 13C NMR
(DMSO-d6) δ 162.44, 161.97, 161.06, 152.58, 134.09,
132.92, 132.80, 127.73, 127.49, 127.15, 124.51, 123.90,
109.89, 66.19. Anal. (C14H12N4O2S) C, H, N.
2.1.15. (2,4-Diaminoquinazoline-6-yl)methyl 2-furoate (15)
Compound 15 was prepared as described for 3 using
the Quest 210 organic synthesizer starting with 2-furoic
acid (177 mg, 1.58 mmol). The reaction was allowed to pro-
ceed for 2 days before it was filtered and worked-up as de-
scribed above for 3 yielding 25 mg (17% over two steps):
1H NMR (DMSO-d6) δ 8.40–8.38 (m, 1H, ArH), 8.06 (d,
J = 1.65 Hz, 1H, H-5), 7.67–7.64 (m, 1H, ArH), 7.56 (dd,
J = 8.58, 1.98 Hz, 1H, H-7), 7.50–7.48 (m, 1H, ArH), 7.31
(br s, 2H, NH2), 7.20 (d, J = 8.58 Hz, 1H, H-8), 6.05
(br s, 2H, NH2), 5.26 (s, 2H, CH2); 13C NMR (DMSO-d6)
δ 162.44, 161.97, 161.06, 152.58, 134.09, 132.92, 132.80,
127.73, 127.49, 127.15, 124.51, 123.90, 109.89, 66.19. Anal.
(C14H12N4O3·0.25H2O) C, H, N.
2.3. Docking and scoring
2.1.16. (2,4-Diaminoquinazoline-6-yl)methyl 3-furoate (16)
Compound 16 was prepared as described for 1 using
2-furoic acid (177 mg, 1.58 mmol). The reaction was allowed
to proceed for 2 days before it was filtered and worked-up
as described above for 1 yielding 30 mg (20% over two
All selected compounds were built and geometry opti-
mized using InsightII (MSI, San Diego). Torsions defining
the orientation of the ester linkage and the hydrophobic sub-
stituents were flexible in the ligands, while the geometry
of the 2,4-diaminoquinazoline moiety was fixed in a planar
structure. Partial charges of the proteins were assigned from
the GROMOS force field (van Gunsteren and Berendsen,
1987), which also was used in the subsequent molecular dy-
namics simulations. The partial charges of the ligands were
1
steps): H NMR (DMSO-d6) δ 8.05 (d, J = 1.65 Hz, 1H,
H-5), 7.96–7.95 (m, 1H, ArH), 7.56 (dd, J = 8.58, 1.65 Hz,
1H, H-7), 7.35–7.34 (m, 3H, NH2 + ArH), 7.20 (d, J =
8.58 Hz, 1H, H-8), 6.69–6.67 (m, 1H, ArH), 6.06 (br s,

48
M. Graffner-Nordberg et al. / European Journal of Pharmaceutical Sciences 22 (2004) 43–53
derived using semi-empirical quantum mechanical calcula-
tions (Marelius et al., 1998a). The energy grids were calcu-
lated using AutoGrid. The grid consisted of 120×120×120
grid points with a spacing of 0.20 Å centered at the active
site of the protein structures, which had the co-crystallized
inhibitors and waters removed. Two crystal structures of hD-
HFR (1HFR (Cody et al., 1998) and 1DLS (Lewis et al.,
1995) modeled from L22Y to wild type) and two of pcD-
HFR (1DYR (Champness et al., 1994) and 3CD2 (Cody
et al., 1999)) were used as models for the target proteins.
Trimethoprim, which is the inhibitor found in the structure
of pcDHFR (1DYR), was used as a reference compound in
order to validate whether the program was able to repro-
duce the experimental structure of the complex. For each
combination of ligand and protein AutoDock was run using
50 Lamarckian genetic algorithm runs (Morris et al., 1998)
with 1 × 106 energy evaluations. The remaining parameters
used by the genetic algorithm were set to the default values
given by AutoDock.
Fig. 2. Trimethoprim in complex with pcDHFR. The position of the
inhibitor as determined by X-ray crystallography is shown in cyan (1DYR).
In orange, the binding conformation as predicted by AutoDock.
2.4. Inhibition of dihydrofolate reductase
After docking, the resulting structures from each run were
sorted into clusters using a 1.0 Å tolerance all-atom root
mean square deviation (RMSD) from the lowest energy
structure. All high-ranked clusters were visually inspected
to make sure that the ligands were docked in a reasonable
orientation in the active site. Trimethoprim was docked with
a root mean square deviation of 1.5 Å compared to the ex-
perimental structure. It was mainly the orientation of the
methoxy groups that contributed to this relatively large de-
viation, but the position of the docked ligand was satisfac-
torily reproduced (Fig. 2).
The compounds 1–16, and the reference compounds
shown in Fig. 1, were evaluated for their ability to inhibit
dihydrofolate reductase from P. carinii (pcDHFR) and rat
liver (rlDHFR). The methodologies of the assay are de-
scribed elsewhere (Broughton and Queener, 1991; Chio and
Queener, 1993). All results are presented as IC50-values in
Table 1. For comparison, the IC50-values for inhibition of
Toxoplasma gondii DHFR (tgDHFR) and Mycobacterium
avium (mavDHFR), respectively, are shown.
Table 1
Inhibition concentrations (IC50, ␮M) of DHFR from P. carinii, T. gondii, and rat liver and selectivity ratios
Compound
IC50 (␮M)
Selectivity indexa
pcDHFR
pcDHFR
rlDHFR
tgDHFR
mavDHFR
tgDHFR
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
0.11
0.85
0.49
0.49
5.8
1.4
0.69
11
–
0.38
1.6
0.52
0.27
1.5
0.093
0.43
0.67
0.21
4
0.13
0.11
19
0.17
0.11
0.73
0.19
0.25
0.35
0.78
0.52
0.39
0.0044
0.008
121
0.1
0.51
1.8
0.67
6.4
0.078
–
1
1.0
3.6
0.36
0.32
1.5
0.05
0.056
0.59
0.10
0.053
0.10
0.18
0.1
0.85
0.51
1.4
0.43
0.69
0.09
0.16
1.7
0.93
0.84
0.37
0.31
0.63
0.042
0.65
8.3
0.52
0.61
1.2
1.5
1.9
1.8
2.4
2.1
3.9
0.26
0.80
45
0.31
0.17
1.73
0.33
0.18
0.62
0.13
0.13
0.19
0.33
0.25
0.099
0.017
0.01
2.7
ND
0.29
0.46
0.37
0.93
0.23
0.18
0.53
0.64
0.13
0.19
4.5
4.3
0.98
0.60
0.034
0.042
27
MGN103
PTXb
TMQb
TMPb
–
a
Defined as the ratio IC50(rlDHFR)/IC50(pcDHFR) or IC50(rlDHFR)/IC50(tgDHFR).
Data taken from Broughton and Queener (1991), and Chio and Queener (1993).
b

M. Graffner-Nordberg et al. / European Journal of Pharmaceutical Sciences 22 (2004) 43–53
49
NH2
N
NH2
N
O
O
K2CO3
DMF
N
Br
N
O
Ar
HO
Ar
+
H2N
H2N
Ar =
1
2
3
4
5
9
6
10
14
8
7
CH3
12
11
H3CO
OCH3
I
CH3
13
15
16
S
O
S
O
Scheme 1.
3. Results and discussion
as two inorganic bases (sodium and cesium carbonate). The
best results and a full conversion of the iodo-compound 9
was accomplished after 120 s at 140 ◦C using Pd(PPh3)2Cl2
as a pre-catalyst with sodium carbonate as a base in
DME/H2O/EtOH (7:3:2) (Scheme 2). After microwave irra-
diation for 80 s, at the same reaction temperature, the start-
ing aryl iodide (9) still remained as deduced from LC–MS
analysis. Yields of 35% (10) and 40% (11) were isolated.
Ester hydrolysis constituted the major side reaction. It is no-
tably that no conversion occurred in the reactions performed
in DMF irrespective of the catalyst employed.
3.1. Chemistry
The target compounds 1–2, 5, 9, 13, and 16, were pre-
pared in DMF by direct displacement of the bromide in
2,4-diamino-6-bromomethylquinazoline (Graffner-Nordberg
et al., 2000) using the appropriate carboxylic acid and
potassium carbonate as a base (Scheme 1). The syntheses
of the esters 3–4, 6–8, 12, and 14–15 were conducted in
the Quest 210 organic synthesizer. The Suzuki-couplings
with the iodo-compound 9 and phenylboronic acid,
or 2,6-dimethoxyphenylboronic acid, providing the es-
ters 10–11, respectively, were conducted in a Smith
SynthesizerTM single mode microwave cavity. For the opti-
mization of the appropriate reaction conditions the ‘Smith
reaction kit’ for the Suzuki couplings was used. The cat-
alysts Pd(PPh3)2Cl2, Pd(OAc)2, and Hermanns catalyst
(trans-di-␮-acetobis[2-(di-o-tolylphosphino)benzyl]dipalla-
dium(II)), respectively, were examined in the synthesis
of 10. Two different solvent systems were employed
[DME/H2O/EtOH (7:3:2), and DMF, respectively], as well
3.2. Docking and scoring
A set of structurally diverse esters comprising various aro-
matic scaffolds were docked and scored using AutoDock 3.0
(Morris et al., 1998). The human reductase (hDHFR) was
used in the docking experiments since high resolution 3D
structures of inhibitor/enzyme complexes are available. The
human reductase exhibits a high homology with rat liver
DHFR (Wang et al., 2001). The estimated free energies of
binding for the highest ranked structure clusters obtained by
R
NH2
N
O
NH2
O
R
Pd(PPh3)2Cl2, Na2CO3,
DME/H2O/EtOH (7:3:2),
(HO)2B
R
O
N
O
N
+
Microwaves
140 ˚C, 2 min
N
H2N
N
I
R
9
10: R = H
11: R = OCH3
R = H alt. OCH3
Scheme 2.

50
M. Graffner-Nordberg et al. / European Journal of Pharmaceutical Sciences 22 (2004) 43–53
Table 2
Calculated free energies of binding obtained by AutoDock compared to experimental IC50-values
Compound
AutoDocka
pc(1DYR)
Experimentala,b
pcDHFR
LIE
pc(3CD2)
h(1DLS)
h(1HFR)
rlDHFR
pc(1DYR)
h(1DLS)
MGN103
1
2
3
4
5
6
7
8
10
12
13
14
15
16
TMP
−13.1
−14.3
−14.6
−16.0
−15.6
−15.0
−15.0
−15.1
−14.9
−15.6
−13.8
−12.6
−12.6
−12.2
−12.4
−11.2
−12.5
−13.8
−14.0
−15.4
−15.6
−13.4
−14.3
−14.2
−14.0
−14.3
−13.2
−11.9
−11.9
−11.5
−11.6
−10.5
−13.6
−15.2
−14.9
−16.1
−16.0
−13.7
−16.0
−14.8
−15.4
−15.5
−13.7
−13.1
−13.0
−12.9
−12.9
−11.6
−14.2
−14.5
−14.3
−16.1
−15.6
−14.4
−15.0
−14.8
−14.9
−15.1
−13.4
−12.2
−12.2
−11.7
−11.7
−10.6
0.60
0.11
0.85
0.49
0.49
5.8
1.40
0.69
11
0.38
0.52
0.27
1.5
0.39
0.09
0.43
0.67
0.21
4
0.13
0.11
19
0.11
0.19
0.25
0.35
0.78
0.52
121
−4.5
0.4
−5.3
−7.0
0.5
0.7
−7.1 0.3
−4.9
−5.2
−4.6
0.3
0.3
0.3
−5.9
−6.0
−5.3
0.5
0.5
0.6
4.3
0.98
27
a
Numbers in bold indicate the five highest ranked compounds in each column.
IC50-values (␮M).
b
AutoDock are summarized in Table 2. As apparent from the
table the magnitude of the binding free energies is overesti-
mated (−16 to −11 kcal/mol), which most likely is attributed
to an effect from an unbalanced scoring function. However,
a comparison of the relative results is still allowed. Although
almost all ligands were predicted to bind slightly stronger to
hDHFR than to pcDHFR, the scores obtained could not re-
veal any significant selectivity among the compounds in the
series. When comparing the scores obtained from two dif-
ferent DHFR structures from the same species [1HFR (Cody
et al., 1998) versus 1DLS (Lewis et al., 1995) modeled from
L22Y to wild type, and two of pcDHFR (1DYR (Champness
et al., 1994) versus 3CD2 (Cody et al., 1999))] the discrep-
ancy in predicted affinity was in the same magnitude as the
difference in affinity when comparing structures from the
different species. This implies that the predicted selectivity
was smaller than the method allowed estimating. However,
the experimental data verified that the selectivity is very low
and that a majority of the compounds have slightly higher
potency as inhibitors of rlDHFR.
Most compounds that gave the highest scores when
docked into the human enzyme were also the best scor-
ing compounds using structures of pcDHFR (Table 2).
Moreover, the experimental data obtained for the top five
compounds demonstrate that three of them also are among
the five most potent inhibitors of rlDHFR. Similarly, for
pcDHFR, three out of the five best scoring compounds
were among the five most potent inhibitors, as found by
experiment. Surprisingly, AutoDock did not rank the most
active pcDHFR inhibitor (1) very high. When focusing on
the lowest ranked compounds the docking and scoring pro-
cedure correctly predicted that TMP has the lowest affinity
to DHFR among all docked compounds. The potency of
compound 5 and 8, as well as the relatively low inhibitory
activity of 6 in the pcDHFR assay, was overestimated by
AutoDock using the current protocol. Thus, AutoDock was
reasonably good at selecting the most potent DHFR in-
hibitors, but less successful when picking the least potent
ones. A more detailed analysis does not seem meaningful,
since both the activity and selectivity of the compounds fall
into a very narrow range.
Five of the docked compounds were also selected for
studies using the more time consuming linear interac-
tion energy method, which is based on energy averages
from molecular dynamics simulations (Åqvist et al., 1994;
Marelius et al., 1998c). This method, which does take into
account explicit waters and flexibility of the protein, has pre-
viously been used to study affinity and selectivity of DHFR
inhibitors (Marelius et al., 1998a; Graffner-Nordberg et al.,
2000, 2001). The calculated binding free energies obtained
from the LIE method also predicted low selectivity of the
compounds (hDHFR/pcDHFR). None of the modeled com-
pounds had a calculated difference in binding free energy
between hDHFR and pcDHFR that was greater than ap-
proximately 1 kcal/mol, in agreement with the experimental
data. (A factor of 10 in binding constant corresponds to
1.36 kcal/mol in binding free energy at 298 K.) Considering
that the selectivity is very low, of the same size as the errors
in the simulations, the predictions by LIE give a qualita-
tively correct picture. Compound 6 represents the only case
where the difference in calculated affinity for pcDHFR and
hDHFR was significantly greater than the errors were. Here
the measured preference for the human reductase by a fac-
tor of 10 is also reproduced, both by LIE and AutoDock.
The absolute values of the binding free energies should be
interpreted with caution. The constant γ in the LIE equa-
tion has not been calibrated to reproduce the experimental
binding energies for this protein-ligand system since only

M. Graffner-Nordberg et al. / European Journal of Pharmaceutical Sciences 22 (2004) 43–53
51
rotation was not productive (cf. 10 and 11), suggesting that
a planar system is preferable. The 2ꢀ-biphenylic compound
(7) exhibited only a moderate potency (IC50 = 0.69 ␮M),
IC50-values and not Ki-values are available (Åqvist and
Marelius, 2001). However, as in the case of the results from
AutoDock it is still relevant to compare relative affinities.
and was approximately equipotent to MGN103 (IC50
=
0.60 ␮M). However, we felt prompted to further exploit the
2-position, which is in a close proximity to the ester bond of
the inhibitor. It was anticipated that proper functionalizations
might later allow for control of the rate of enzymatic hydrol-
ysis, an important issue to address in the final fine-tuning of
the systems. Displacement of the biphenyl structure for an-
other very common privilege structure, a diphenylmethane
group (Hajduk et al., 2000), afforded 6, that was found to
be less active than 7. Apparently, the insertion of methylene
group in 7 to produce a compound with higher flexibility
was not well tolerated by pcDHFR. Considering the other
diphenylmethyl derivative, compound 5, this ester is a poor
inhibitor of the DHFRs, although still better than the corre-
sponding benzyl ester, lacking one of the two phenyl rings
(Graffner-Nordberg et al., 2001).
3.3. Dihydrofolate reductase inhibition
The impact of various linkers connecting the two ring sys-
tems has previously been studied thoroughly (Blaney et al.,
1984). The lipophilic parts of the ester-based inhibitors 1–16
comprise in this study aromatic systems of different sizes
and topography. These systems will serve as scaffolds for
further synthetic modifications. We first assessed the influ-
ence of small methyl groups in the 3- and 5-positions of
the phenyl ring of MGN103 (Fig. 1). This modification ren-
dered an inhibitor (12) that became slightly more potent than
the parent compound and with approximately the same in-
hibitory effect as the corresponding methoxy-substituted in-
hibitors (Graffner-Nordberg et al., 2001). Since both alkoxy
and methyl groups polarize aromatic systems, we found it
relevant to introduce the heterocyclic thiophene and furan
nucleus, both with dipoles directed towards the heteroatoms
as net effects, as substitutes for the phenyl ring. A compar-
ison of the four heterocycles (13–16) revealed a dramatic
impact on the inhibitory capacity. The high activity of the
2-thienyl derivative (0.27 ␮M) as compared to the corre-
sponding 2-furyl compound (4.3 ␮M) is the most striking
observation. The fact that a substitution of the phenyl group
for a 2-thienyl delivers a twice as active inhibitor while a
substitution for the 2-furyl is deleterious is not easy to ra-
tionalize. More potent inhibitors than the parent compound
MGN103, could also be obtained by an enlargement of the
system. Thus, the best inhibitor in the series was obtained by
displacement of the phenyl ring for a 1-naphthyl group (1).
Anchoring of the ester group to a 2-naphthyl group (2) re-
sulted in poorer inhibition. Displacements of the 1-naphthyl
moiety in 1 for a phenanthrene or an anthracene tricyclic
ring system, 3 and 4, respectively, did not lead to further
improvements. It is notable though, that while compound 3
and 4 exhibited the same IC50-values in the pcDHFR assay,
the phenanthryl derivative (3) was in fact a relatively poor
inhibitor of the mammalian reductase [IC50(rlDHFR) =
0.67 ␮M] and demonstrated a slight preference for the fungal
enzyme. Rosowsky et al. (Rosowsky et al., 1999) previously
reported the impressive selectivity index of 21 (pcDHFR:
0.21 ␮M; rlDHFR: 4.4 ␮M) with a related nitrogen function-
alized tricyclic iminostilbene, where a propensity to adopt a
puckered orientation was convincingly suggested to account
for the high selectivity. Biphenyls were addressed next as
scaffolds. A comparison of the impact of phenyl groups at-
tached to the 2, 3 and 4-positions of MGN103, revealed an
acceptance for an aromatic ring in the 2-position (7) and
3-position (10), but not for a substitution in the 4-position
(8). The 4ꢀ-biphenyl compound (8) was even less tolerated
by rlDHFR and exhibited the highest selectivity index within
the series. Introduction of methoxy groups adjacent to the
pivotal bond of the 3ꢀ-phenyl substituted derivative to restrict
3.4. Molecular dynamics
Both with regard to the ranking of different inhibitors,
as well as the selectivity, there is a reasonable correlation
between the results obtained from AutoDock and the LIE
method. However, the predicted ranking of the affinities
disagrees with experiment in some cases, which may reflect
an inexact docking mode. AutoDock docks the flexible lig-
ands into the active site of rigid protein structures and thus,
reorientation of side chains upon ligand binding cannot be
taken into account. Furthermore, it is also difficult to repro-
duce the effect of explicit water molecules that may bridge
ligand–protein interactions. This problem is of particular
relevance in the present study since antifolates bind partly
via water bridges to the enzyme (Roth, 1986; Poornima
and Dean, 1995; Meiering and Wagner, 1995). The quality
of the scoring function is crucial for correct ranking of the
docking modes as well as the estimation of the binding free
energy. Nevertheless, automated docking is a fast and use-
ful method for screening and affinity predictions on a large
set of compounds. Both LIE, and the scoring function used
by AutoDock, underestimated the relative affinity of the
4ꢀ-biphenyl compound (8) to pcDHFR. Both methods also
seemed to rank the phenanthryl derivative (3) higher than
was found by experiment. Furthermore, a relatively high ac-
tivity of the 1-naphthyl analogue (1) was not reproduced. It
seems likely that with the present docking procedure there
may be a problem in predicting conformational changes
that might occur upon binding of large hydrophobic lig-
ands. Even with the more sophisticated molecular dynamics
technique it can be difficult to model the correct displace-
ments of amino acid side chains and water molecules if the
initially docked structure is non-optimal.
Until high resolution 3D-DHFR complexes with ligands
encompassing the ester linkage are available as starting
structures for docking programs and MD simulations,

52
M. Graffner-Nordberg et al. / European Journal of Pharmaceutical Sciences 22 (2004) 43–53
References
Allegra, C.J., Chabner, B.A., Tuazon, C.U., Ogata-Agakaki, D., Baird,
B., Drake, J.C., Simmons, J.T., Lack, E.E., Shelhamer, J.H., Balis, F.,
Walker, R., Kovacs, J.A., Lane, H.C., Masur, H., 1987a. Trimetrexate
for the treatment of Pneumocystis carinii pneumonia in patients with
the acquired immunodeficiency syndrome. N. Engl. J. Med. 317, 978–
985.
Allegra, C.J., Kovacs, J.A., Drake, J.C., Swan, J.C., Chabner, B.A., Ma-
sur, H., 1987b. Activity of antifolates against Pneumocystis carinii
dihydrofolate reductase and identification of a potent new agent. J.
Exp. Med. 165, 926–931.
Anonymous, 1994. News. FDA approves trimetrexate as a second line
therapy for Pneumocystis carinii pneumonia. Am. J. Hosp. Pharm.
51, 591–592.
Åqvist, J., Marelius, J., 2001 The linear interaction energy method for
predicting ligand binding free energies. In: Reddy, M.R., Erion, M.D.
(Eds.), Free Energy Calculations in Rational Drug Design. Kluwer
Academic Publishers, Dordrecht.
Åqvist, J., Medina, C., Samuelsson, J.E., 1994. A new method for pre-
dicting binding affinity in computer-aided drug design. Protein Eng.
7, 385–391.
Fig. 3. Compound 3 docked into the active site of pcDHFR (1DYR) using
AutoDock. The orientation of the ester moiety is not entirely clear, since
the two docked conformations only differ by 0.07 kcal/mol in binding free
energy. The conformation where the carbonyl oxygen is pointing upwards
is predicted to have the lowest energy.
Arozullah, A.M., Yarnold, P.R., Weinstein, R.A., Nwadiaro, N., McIlraith,
T.B., Chmiel, J.S., Sipler, A.M., Chan, C., Goetz, M.B., Schwartz,
D.N., Bennett, C.L., 2000. A new preadmission staging system for pre-
dicting inpatient mortality from HIV-associated Pneumocystis carinii
pneumonia in the early highly active antiretroviral therapy (HAART)
era. Am. J. Respir. Crit. Care Med. 161, 1081–1086.
uncertainties will remain regarding the preferred confor-
mation of the ester group bridging the aromatic parts, as
indicated in Fig. 3.
Bertino, J.R., Sawacki, W.L., Moroson, B.A., Cashmore, A.R., El-
slager, E.F., 1979. 2,4-Diamino-5-methyl-6-[(2,4,5-trimethoxyanilino)-
methyl]quinazoline (TMQ), a potent non-classical folate antagonist
inhibitor-I. Biochem. Pharmacol. 28, 1983–1987.
Blaney, J.M., Hansch, C., Silipo, C., Vittoria, A., 1984. Structure-activity
relationships of dihydrofolate reductase inhibitors. Chem. Rev. 84,
333–407.
Bodor, N., Buchwald, P., 2000. Soft drug design: general principles and
recent applications. Med. Res. Rev. 20, 58–101.
Broughton, M.C., Queener, S.F., 1991. Pneumocystis carinii dihydrofolate
reductase used to screen potential antipneumocystis drugs. Antimicrob.
Agents Chemother. 35, 1348–1355.
4. Conclusion
Even though the objective of the present program is to
achieve a favorable toxicity profile by applying the soft
drug concept, it is still highly desirable to optimize the
pcDHFR/hDHFR ratio to suppress host toxicity at the site
of administration. The most potent of the new ester-based
DHFR-inhibitors, the 1-naphthyl derivative (1) exhibits an
IC50-value of 110 nM, and is less active than trimetrexate
with an IC50-value of 42 nM. However, the selectivity in-
dex is improved (0.85 versus 0.19 for TMQ). In the group
of the most potent inhibitors, the phenanthryl derivative (3)
exhibits the highest selectivity index, which is still far lower
than what is observed with TMP that is used in clinic for
treatment of PCP (1.4 versus 4.5 for TMP), in the present
assay systems. It is notably though that the new phenan-
thryl ester exerts a 50 times higher inhibitory activity than
TMP. Regarding the computational approaches used for pre-
diction, the calculated results fell into a very narrow range
with respect to both affinity and selectivity. The predictions
were, with a few exceptions, in qualitatively good agreement
with experiments, suggesting that these scoring programs
can serve as valuable research tools and guide for further
selection of compounds as targets for synthesis.
Champness, J.N., Achari, A., Ballantine, S.P., Bryant, P.K., Delves,
C.J., Stammers, D.K., 1994. The structure of Pneumocystis Carinii
dihydrofolate reductase to 1.9
924.
Å resolution. Structure 2, 915–
Chio, L.-C., Queener, S.F., 1993. Identification of highly potent and se-
lective inhibitors of Toxoplasma gondii dihydrofolate reductase. An-
timicrob. Agents Chemother. 37, 1914–1923.
Clissold, S.P., Heel, R.C., 1984. Budesonide: a preliminary review of its
pharmacodynamic properties and therapeutic efficacy in asthma and
rhinitis. Drugs 28, 485–518.
Cody, V., Galitsky, N., Luft, J.R., Pangborn, W., Blakley, R.L., Gangjee,
A., 1998. Comparison of ternary crystal complexes of F31 variants of
human dihydrofolate reductase with NADPH and a classical antitumor
furopyrimidine. Anticancer Drug Des. 13, 307–315.
Cody, V., Galitsky, N., Rak, D., Luft, J.R., Pangborn, W., Queener, S.F.,
1999. Ligand-induced conformational changes in the crystal structures
of Pneumocystis carinii dihydrofolate reductase complexes with folate
and NADP+. Biochemistry 38, 4303–4312.
Duch, D.S., Edelstein, M.P., Bowers, S.W., Nichol, C.A., 1982.
Biochemical and chemotherapeutic studies on 2,4-diamino-6-(2,5-
dimethoxybenzyl)-5-methylpyrido[2,3-d]pyrimidine (BW301U),
novel lipid-soluble inhibitor of dihydrofolate reductase. Cancer Res.
42, 3987–3994.
A
Acknowledgements
We thank the Swedish Research Council (VR) and the
Swedish Foundation for Strategic Research (SSF) for finan-
cial support.
Elslager, E.F., Davoll, J., 1974. Synthesis of fused pyrimidines as fo-
late antagonists. In: Castle, R.N., Townsend, L.B. (Eds.), Lectures in
Heterocyclic Chemistry. Hetero Corp., Orem, UT.

M. Graffner-Nordberg et al. / European Journal of Pharmaceutical Sciences 22 (2004) 43–53
53
Elslager, E.F., Johnson, J.L., Werbel, L.M., 1983. Folate antagonists.
20. synthesis, antitumor, and antimalarial properties of trimetrexate
and related 6-[(phenylamino)methyl]-2,4-quinazolindiamines. J. Med.
Chem. 26, 1753–1760.
Lewis, W.S., Cody, V., Galitsky, N., Luft, J.R., Pangborn, W.,
Chunduru, S.K., Spencer, H.T., Appleman, J.R., Blakley, R.L., 1995.
Methotrexate-resistant variants of human dihydrofolate reductase with
substitutions of leucine 22: kinetics, crystallography, and potential as
selectable markers. J. Biol. Chem. 270, 5057–5064.
Lin, J.T., Cashmore, A.R., Baker, M., Dreyer, R.N., Ernstoff, M., Marsh,
J.C., Bertino, J.R., Whitfield, L.R., Delap, R., Grillo-Lopez, A.,
1987. Phase I studies with trimetrexate: clinical pharmacology, an-
alytical methodology, and pharmacokinetics. Cancer Res. 47, 609–
616.
Marelius, J., Graffner-Nordberg, M., Hansson, T., Hallberg, A., Åqvist,
J., 1998a. Computation of affinity and selectivity: binding of
2,4-diaminopteridine and 2,4-diaminoquinazoline inhibitors to dihy-
drofolate reductases. J. Comput. Aided Mol. Des. 12, 119–131.
Marelius, J., Hansson, T., Åqvist, J., 1998c. Calculation of ligand binding
free energies from molecular dynamics simulations. Int. J. Quant.
Chem. 69, 77–88.
Marelius, J., Kolmodin, K., Feierberg, I., Åqvist, J., 1998b. Q: an MD
program for free energy calculations and empirical valence bond
simulations in biomolecular systems. J. Mol. Graph. Model. 16, 213–
225.
Falloon, J., Allegra, C.J., Kovacs, J., O’Neill, D., Ogata-Arakaki, D.,
Feuerstein, I., Polis, M., Davey, R., Lane, H.C., LaFon, S., Rogers,
M., Zunich, K., Turlo, J., Tuazon, C., Parenti, D., Simon, G., Masur,
H., 1990. Piritrexim with leucovorin for the treatment of Pneumocystis
Pneumonia (PCP) in AIDS patients. Clin. Res. 38, 361A.
Fischl, M.A., Dickinson, G.M., La Voie, L., 1988. Safety and efficiency
of sulfamethoxazole and trimethoprim chemoprophylaxis for Pneumo-
cystis carinii pneumonia in AIDS. J. Am. Med. Assoc. 105, 45–48.
Glatt, A.E., Chirgwin, K., 1990. Pneumocystis carinii pneumonia in human
immunodeficiency virus-infected patients. Arch. Intern. Med. 150,
271–279.
Golden, J.A., Chernoff, D., Hollander, H., Feigal, D., Conte, J.E., 1989.
Prevention of Pneumocystis carinii by inhaled pentamidine. Lancet 1,
654–657.
Gordin, F.M., Simon, G.L., Wofsy, C.B., Mills, J., 1984. Adverse reac-
tions to trimethoprim-sulfamethoxazole in patients with the acquired
immunodeficiency syndrome. Ann. Intern. Med. 100, 495–499.
Graffner-Nordberg, M., Kolmodin, K., Åqvist, J., Queener, S.F., Hallberg,
A., 2001. Design, synthesis, computational prediction and biological
evaluation of ester soft drugs as inhibitors of dihydrofolate reductase
from Pneumocystis carinii. J. Med. Chem. 44, 2391–2402.
Graffner-Nordberg, M., Marelius, J., Ohlsson, S., Persson, Å., Swedberg,
G., Andersson, P., Andersson, S.E., Åqvist, J., Hallberg, A., 2000.
Computational predictions of binding affinities to dihydrofolate reduc-
tase: synthesis and biological evaluation of methotrexate Analogues.
J. Med. Chem. 43, 3852–3861.
Grivsky, E.M., Lee, S., Sigel, C.W., Duch, D.S., Nichol, C.A.,
1980. Synthesis and antitumor activity of 2,4-diamino-6-(2,5-dime-
thoxybenzyl)-5-methylpyrido[2,3-d]pyrimidine. J. Med. Chem. 23,
327–329.
Hajduk, P.J., Bures, M., Praestgaard, J., Fesik, S.W., 2000. Privileged
molecules for protein binding identified from NMR-based screening.
J. Med. Chem. 43, 3443–3447.
Masur, H., 1992. Prevention and treatment of Pneumocystis pneumonia.
N. Engl. J. Med. 327, 1853–1860.
Masur, H., Polis, M.A., Tuazon, C.V., Ogota, A.D., Kovacs, J.A., Katz, D.,
Hilt, D., Simmons, T., Feuerstein, I., Lundgren, B., Lane, H.C., Chab-
ner, B.A., Allegra, C.J., 1993. Salvage trial of trimetrexate-leucovorin
for the treatment of cerebral toxoplasmosis. J. Infect. Dis. 167, 1422–
1426.
McKenzie, R., Travis, W.D., Dolan, S.A., 1991. The causes of death in
patients with human immunodeficiency virus infections: a clinical and
pathological study with emphasis on the role of pulmonary diseases.
Medicine (Baltimore) 70, 326–343.
Meiering, E.M., Wagner, G., 1995. Detection of long-lived bound wa-
ter molecules in complex of human dihydrofolate reductase with
methotrexate and NADPH. J. Mol. Biol. 247, 294–308.
Miller, R.F., Mitchell, D.M., 1995. AIDS and the lung. Update 1995: 1
Pneumocystis carinii pneumonia. Thorax 50, 191–200.
Mills, J., 1986. Pneumocystis carinii and Toxoplasma gondii infections
in patients with AIDS. Rev. Infect. Dis. 8, 1001–1011.
Monk, J.P., Benfield, P., 1990. Inhaled pentamidine. An overview of its
pharmacological properties and a review of its therapeutic use in
pneumocystis carinii pneumonia. Drugs 39, 741–756.
Morris, G.M., Goodsell, D.S., Halliday, R.S., Huey, R., Hart, W.E.,
Belewm, R.K., Olson, A.J., 1998. Automated docking using a lamar-
ckian genetic algorithm and empirical binding free energy function.
J. Comput. Chem. 19, 1639–1662.
Pearson, R.D., Hewlett, E.L., 1985. Pentamidine for the treatment of
Pneumocystis carinii pneumonia and other protozoal diseases. Ann.
Intern. Med. 103, 782–786.
Hirschel, B., Lazzarin, A., Chopard, P., 1991. A controlled study of
inhaled pentamidine for primary prevention of Pneumocystis carinii
pneumonia. N. Engl. J. Med. 324, 1079–1083.
Johansson, S.-Å., Andersson, K.-E., Brattsand, R., Gruvstad, E., Hedner,
P., 1982. Topical and systemic glucocorticoid potencies of budesonide
and beclomethasone dipropionate in man. Eur. J. Clin. Pharmacol. 22,
523–529.
Kaplan, J.E., Hanson, D., Dworkin, M.S., Frederick, T., Bertolli, J.,
Lindegren, M.L., Holmberg, S., Jones, J.L., 2000. Epidemiology of
human immunodeficiency virus-associated opportunistic infections in
the united states in the era of highly active antiretroviral therapy. Clin.
Infect. Dis. 30 (Suppl. 1), S5–S14.
Poornima, C.S., Dean, P.M., 1995. Hydration in drug design. 3. Conserved
water molecules at the ligand-binding sites of homologous proteins.
J. Comput. Aided Mol. Des. 9, 521–531.
Rosowsky, A., Cody, V., Galitsky, N., Fu, H., Papoulis, A.T., Queener, S.F.,
1999. Structure-based design of selective inhibitors of dihydrofolate
reductase: synthesis and antiparasitic activity of 2,4-diaminopteridine
analogues with a bridged diarylamine side chain. J. Med. Chem. 42,
4853–4860.
Kovacs, J.A., Allegra, C.A., Swan, J.C., Drake, J.C., Parillo, J.E., Chabner,
B.A., Masur, H., 1988. Potent antipneumocystis and antitoxoplasma
activities of piritrexim, a lipid-soluble antifolate. Antimicrob. Agents
Chemother. 32, 430–433.
Kovacs, J.A., Hiemenz, J.W., Macher, A.M., Stover, D., Murray, H.W.,
Shelhamer, J., Lane, H.C., Urmacher, C., Honig, C., Longo, D.L.,
Parker, M.M., Natanson, C., Parillo, J.E., Fauci, A.S., Pizzo, P.A., Ma-
sur, H., 1984. Pneumocystis carinii pneumonia: a comparison between
patients with the acquired immunodeficiency syndrome and patients
with other immunodeficiencies. Ann. Intern. Med. 100, 663–671.
Lee, F.S., Warshel, A., 1992. A local reaction field method for fast evalu-
ation of long-range electrostatic interactions in molecular simulations.
J. Chem. Phys. 97, 3100–3107.
Roth, B., 1986. Design of dihydrofolate reductase inhibitors from X-ray
crystal structures. Fed. Proc. 45, 2765–2772.
Roudier, C., Caumes, E., Rogeauz, O., Bricaire, F., Gentilini, M., 1994.
Adverse cutanous reactions to trimethoprim-sulfamethoxazole in pa-
tients with the acquired immunodeficiency syndrome and Pneumocys-
tis carinii pneumonia. Arch. Dermatol. 30, 1383–1386.
Sattler, F.R., Allegra, C.J., Verdegem, T.D., 1990. Trimetrexate-leucoverin
dosage evaluation study for the treatment of Pneumocystis carinii
pneumonia. J. Infect. Dis. 161, 91–96.
Leoung, G.S., Feigal, D.W., Montgomery, A.B., Corkery, K., Wardlaw, L.,
Adams, M., Busch, D., Cordon, S., Jacobson, M.A., Volberding, P.A.,
Abrams, D., 1990. Aerosolized pentamidine for prophylaxis against
Pneumocystis carinii pneumonia. N. Engl. J. Med. 323, 769–775.

54
M. Graffner-Nordberg et al. / European Journal of Pharmaceutical Sciences 22 (2004) 43–53
Schliep, T.C., Yarrish, R.L., 1999. Pneumocystis carinii pneumonia.
Semin. Respir. Infect. 14, 333–343.
Wang, Y., Bruenn, J.A., Queener, S.F., Cody, V., 2001. Isolation of rat
dihydrofolate reductase gene and characterization of recombinant en-
zyme. Antimicrob. Agents Chemother. 45, 2517–2523.
Weir, E.C., Cashmore, A.R., Dreyer, R.N., Graham, M.L., Hsiao, N.,
Moroson, B.A., Sawacki, W.L., Bertino, J.R., 1982. Pharmacology
and toxicity of a potent “nonclassical” 2,4-diamino quinazoline folate
antagonist, trimetrexate, in normal dogs. Cancer Res. 42, 1696–1702.
van Gunsteren, W.F., Berendsen, H.J.C., 1987. Groningen Molecular Sim-
ulation (GROMOS) Library Manual, Biomos B.V., Groningen.
Walker, R.E., Masur H., 1994. Current regimens of therapy and prophy-
laxis. In: Walzer, P.D. (Ed.), Pneumocystis carinii Pneumonia. Marcel
Dekker, New York, pp. 439–466.